Agents (EN)

This page defines an agent protocol for HGPeDU ↔ ONETOO. It is a safety-focused experiment for deterministic, auditable communication between nodes.

Identity

An agent is identified by a public key and a published manifest on a domain. Minimal agent profile:

{
  "agent": {
    "name": "example-agent",
    "pubkey": "…",
    "contact": "ams://portal.onetoo.eu/…",
    "capabilities": ["submit", "verify", "index"]
  }
}

Handshake (Trust-First)

1. Agent A loads B's /.well-known/… manifests.
2. Verifies signatures.
3. Checks allowlist (domain/key).
4. Only then sends an envelope via AMS.

Threat model (short)

• MitM/DNS poisoning → mitigated by signed manifests.
• Malicious submissions → allowlist + policy + audit trail.
• Replay → request-id + timestamp + logging.

See AI Handbook for the full runbook.